When creating a new compression spec, my team always sacrificed a slight bit of the video bit rate to ensure the audio bit rate was high enough to not be annoying. For example, doubling 64k to 128k makes significant improvement to audio, but does not really help/hurt the video. Take a low bit rate video and have one version with decent audio and another with bad audio, and people will say that the video with bad audio is worse even though the video quality did not change.

I'm not talking about that compression, I'm talking about dynamic range compression.